Well, as you all know i have upto now be mostly a proponent of the keep non-free camp, because, altough i fully would prefer every software in debian to be free, i feel that this is not really yet the time for it.
Now, there are three catgoeries of non-free packages, and Matt mostly mentions one of them. The categories are : 1) non-free documentation, including non-free documentation of free packages and non-free documentation currently hiding in main. Maybe non-free data can be added here too, but i guess the situation is different, and maybe non-free data deserves a category of its own. 2) non-free hardware drivers. 3) other assorted non-free packages. Not sure about data, but data can most easily be packaged in a different format anyway, to at least suppress the redundancy of having it two times, once in source package, once in the 'binary' one, maybe a new dpkg support who knows to get data (from various third parties) and install it in system conformant place would be a solution. Anyway, the case of non-free documentation of free software is awkward. Even if the documentation is non-free, it is often the case that it provides a useful addition to a free package, and that modification are most easily accepted upstream. The only case where this does cause a problem is if you are going to make a hostile fork of the the free package. The case of non-free hardware i will go back again below, since this is the main point here. The case of assorted non-free packages not falling in the above categories, well, nothing major is left, so let's either remove or reimplement them, or have the upstream authors relicence it. The fact that many of the licences are almost or quasi free could be a reason for keeping them, but often, i agree, they are just being stubborn. That said, it would be best to make this decision a case by case decision. Also, just moving them to a third party non-free.org will probably achieve nothing, so i believe it is not worth it. Now, on the non-free hardware drivers. I have not heard the non-free removal proponent propose anything serious about this, and i think this is the main point of disagreement. On the other hand, these non-free drivers are most often binary-only x86 drivers, and consist of the worst case of non-free stuff there can be, and also often said hardware manufacturer cheat their customer by claiming "linux support" on the package, while it is only linux/x86 support, and often not even the most recent version thereof. To the actual manufacturers benefit, it is not always entirely their fault, they are often not able to easily release the IP under a free licence, because they don't own them themselves, but still there is also clearly a lack of motivation for it. So, if we are going to remove non-free, what will be our stand on hardware drivers ? Do we go as far as publicly state that we are not going to support those ? Do we put all our pressure to fight it and engage in the combat against closed source hardware ? Or do we stay hypocritical and stop distributing the above mentioned software categories and on the other hand silently install those hardware drivers from third sources on our own boxes ? Also, this would mean open support to those projects aiming to replace non-free pieces of low level stuff, like the free bios/of projects, or other free hardware design projects ? But as said, i believe we are not bloody likely to achieve something there just by ourself, at least as long as linus consider binary only modules as acceptable. Anyway, back to Matt's post. On Wed, Mar 03, 2004 at 03:39:18PM -0600, Matt Pavlovich wrote: > The position to remove non-free as integrated part of Debian is a > "technology first", "end user second" position. While the goal for a > 100% free distribution is a great goal, I draw the line when users are > negatively impacted for the sake of the goal. It boils down to -- Who > do you put first-- the softare or the users? A clear example of my > point is with the Nvidia driver. > > What happens when Nvidia releases the "GeForce X" and there is no driver > support in the XFree86 module, but there is in the Nvidia driver? > Debian end user loses. It is not even possible now. Try running 3D accel on a nvidia equiped powerbook. And newer ATI hardware will also come spec-less. > What happens when there is support in the XFree86 driver, but it is for > XFree86 4.4 and their license is still prohibitive? Debian end user > loses. No, because none of the drivers fall under that licence, so it would be easy to port them to a free fork if that ever became needed. > Some argue that distributing the Nvidia driver, and including it in the > X configuration is conforming to Nvidia's evil intentions. I find this > a bit far fetched given the current small number of Linux desktops and > my personal experiences working with Nvidia. I also find it a bit rash > to compare the Nvidia video drivers for Linux to be on the same level of > evilness as a Microsoft. Nvidia integrates with existing (and custom) > kernels and XFree86. Try using them on powerpc, and we will speak again. > *Users do not have to pay extra for the Linux software > *Nvidia allows redistribution of the software > *Users are not forced to run "Nvidia Kernels" > *Users are not forced to run "Nvidia Xwindows" > *Nvidia software is more feature rich and useful to users than the 100% > free alternative > -Supports 3d > -Support Xv > -Supports DVI > -Supports multi-head Doesn't support anything but x86 and is often broken. If they were going to be serious, and provide builds for at least the major agp supporting non-x86 hardware out there, then why not, but they don't care about this. This is probably because their driver is not endian clean, and probably full of unportable x86 assembly. > Nvidia's model is very similar to that of Netscape, and it is difficult > to argue against the positive effect that Netscape had for the free > desktop. Nope it is not. Netscape gave the sources in a free way, while nvidia does not, and don't even distribute powerpc binaries, while it is their second most important market, considering all the apple hardware that uses it. > Debian's position to not include support for the binary-only driver > negatively impacts end users. Having a non-free video driver as the > only piece of non-free software on a system is a great leap from a 100% > non-free system. Yeah, but still, the binary-only driver are the true evil here. Not the one who don't like their software to be used by mass murderers, or the governement, or ... > >From a technical standpoint, I urge people to consider that the > 'Netscape model' is a great way for 3rd party vendors to move from a > 100% closed source model to a free one over time. Debian can be a great > vessel to facilitate these transitions. > > >From a priority standpoint, given the fact that Nvidia is a top 2 video > chip manufacturer, I consider it core to the success of the free desktop > (and to some extent Debian) to put the end user first. And it is because many here, including proponent of the non-free removal probably, think like that, that things will never change. Even worse, the situation is becoming worse today that they where a few years back, where both ATI and matrox where known for giving out specs, where 3Dlabs gave full specs easily, and where nvidia, and its do everything in-house politic was mostly a minor player. But these days are over, both matrox and ATI latest hardware doesn't come with specs, and Nvidia is in the top 2 like you say. So, if we are going to be honest and remove non-free, we should go all the way, and actively fight the binary-only hardware drivers, putting all our influence, and i believe many will hear us over this, in making the hardware manufacturers understand that we are not going to accept this. But then, i have heard no such thing from the non-free removal proponents, and i fear that they are content in a future where there is no free 3D support for linux, and we will have to resort to non-free binary only hardware drivers to have this, and then even, only on x86 boxes, thus killing all chance of other hardware in the linux desktop. Friendly, Sven Luther -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
